International Journal of Engineering & Technology IJET-IJENS Vol: 12 No: 01 57
Improving the Performance of Temperature
Model of Economizer Using Bond Graph and
Genetic Algorithm
A.N. Aziz, P. Siregar, Y.Y. Nazaruddin, and Y. Bindar
accepted as one of global optimization [Link] searches
Abstract—The mathematical model of economizer, based on global optimum solution by imitate evolutionary mechanism
physical laws is derived using bond graph [Link] in nature, like permutation, crossover as well as combination
multi-energy bond graph, which employs mass flow-rate and of chromosome, over specified population through a sort of
enthalpy flow-rate as flow variables, as well as pressure and
[Link] method has been proven successfully to solve
temperature as effort variables, is used in achieving [Link]
heat transfer coefficient of economizer is obtained by using optimization problems where conventional techniques, such as
logarithmic temperature mean difference between flue gas and gradient method, failed to do [Link] employment of GA to
boiler feed water.A modification of overall heat transfer optimize binary distillation unit [11], to control neutralization
coefficient in the form of parameterized polynomial is also done process [9], and to optimize the performance of model
by using the help of genetic algorithm technique.A step predictive controller [10] have demonstrated its superior
simulation of the model at maximum, continuous, and minimum
potentials.A survey paper about utilization of GA in modeling
boiler operating condition demonstrates, that model’s
performance has been improved. and control of combustion process can be found in[6].
This article will outline the idea of developing mathematical
Index Terms—bond graph, genetic algorithm, mathematical models with a couple of parameters for economizer. The
model, heat transfer coefficient mathematical model derivation is guided by bond graph
modeling [Link] unknown parameters will be obtained
using GA techniques.
I. INTRODUCTION The rest of the paper will be organized as [Link] 2
D ue to boiler aging, uncertainties, high nonlinearities, and
un-modeled dynamics, modeling error between first
principle based model and real industrial boiler
briefly describes steam generation process and it
[Link] models developments and parameter
estimation will be discussed in section [Link] 4 elaborates
responsesalways [Link] arise when the requirement the result and its evaluation, and finally, section 5 concludes
is having an adequate model with computational resourceskept with conclusion and suggestion.
at low [Link] problem often appears in model-based
control design. II. SYSTEM DESCRIPTION
The grey-box method can be employed to compromise The investigation will be conducted at a water-tube boiler
these [Link] utilizes either analytical plant model, with system which is fueled by fossil [Link] boiler has 110 ton/h
some unknown parameters based on fundamental laws of capacities that produces steam into header at a specific
physics, and system identification [Link] or pressure and [Link] pressure has to be maintained at
adaptation of parameters are then performed by particular 60 kg/cm2 and the temperature is kept at 460 0C, regardless of
optimization algorithm. the load condition.
In recent years, genetic algorithm (GA) has been widely The heat is supplied from a furnace, which is equipped with
three burners, and reaches the boiler tubes by radiation and
Manuscript received January 13, 2012. This work was supported in part by convection [Link] hot water is boiled through the
Lembaga Penelitian dan Pengabdian Masyarakat (LPPM), Institute Teknologi
Bandung under Riset KK scheme. pipes and converted into steam at the upper end of the
A. N. Aziz is with the Electronics and Instrumentation Research Group, [Link] steam product is also superheated by passing it
Physics Program Study, Faculty of Science and Engineering,Jenderal through the [Link]. 1 shows the boiler cross section
Soedirman University, Jl. Dr. Soeparno 61, 53123 Purwokerto, Indonesia (e-
mail: [Link]@[Link]). under investigation.
P. Siregar is with Instrumentation and Control Research Group, The high-pressure steam from outlet header goes into the
Engineering Physics Program Study, Faculty of Industrial Technology, ITB, Jl main steam header which is connected to the generator,
Ganesha 10 Bandung-40132.
Y. Y. Nazaruddin is with Instrumentation and Control Research Group, turbine, and other processes at the refinery plant.
Engineering Physics Program Study, Faculty of Industrial Technology, ITB, Jl The economizer is equipped with two headers of 165.2 mm
Ganesha 10 Bandung-40132. in outer [Link] water is fed into the lower header and
Y. Bindar is Energy and Processing System of Chemical Engineering,
Faculty of Industrial Technology, ITB, Jl Ganesha 10 Bandung-40132.
the boiler water is supplied from the upper [Link]
126701-5858 IJET-IJENS @ February 2012 IJENS
IJENS
International Journal of Engineering & Technology IJET-IJENS Vol: 12 No: 01 58
structure is made to allow counter flow of the feed [Link] has m&
547 square-meters of surface [Link] is also made of carbon m&
steel material. T1
MSf : m& c p T1
T1 T2
T1 T1 T2
H& H&
H& H& H&
Fig. 2. CETF (Coupling Element for Thermo-Fluid)
A typical schematic diagram of input-process-output and
bond graph model of economizer using CETF as the coupling
element are shown below.
pin Economizer pout
Tin Waterin Waterout Tout
H& in
Teco H&out
m& in &out
m
H& gas → fluid
Fig. 1. Boiler cross section and economizer
Qapplied
III. METHODOLOGY
Fig. 3. Schematic diagram of input-process-output of economizer
A. Bond graph model of economizer
It has been mentioned elsewhere, that pseudo power B. Parameter estimation using GA
variables for thermal and hydraulic systems in process As an optimization tool, GA works with a set of
engineering are temperature (K) plus enthalpy flow (J/s), and solutions, = , , ⋯ , , called population.A population
element, , is called an [Link] each iteration, GA
pressure (P) plus mass flow (kg/s), respectively[12].In case of
economizer, there are assumed two systems involved, i.e.
evaluates the fitness of all the individuals in the population
p1 1
p2
Sf 1
m& 1 m& 1
m& 1 m& 1
T1
MSf : m& he (T1 ) MSf : m& he (T2 )
C
T1 T2 T2
Se 0 1 0 0 1
m& h e (T 1 ) m& h e (T 2 )
H& applied
Sf
Fig.4. Bond graph model of economizer
thermal and [Link] multi-energy bond-graph should and creates new population by performing operations such as
represent the coupling behaviors of thermal and hydraulic combining two individuals (crossover) or changing an
[Link] is done by using CETF (coupling element for individual (mutation).The old population is discarded, so GA
thermo-fluid) as stated elsewhere in[8].It takes the following will start a new process using the new [Link]
form (Fig. 2). iteration is referred to as a generation.
In each generation individuals are selected for reproduction
according to their performance with respect to fitness
126701-5858 IJET-IJENS @ February 2012 IJENS
IJENS
International Journal of Engineering & Technology IJET-IJENS Vol: 12 No: 01 59
[Link] selection is done in such a way that the feed water inlet and outlet temperature, flue gas inlet and
individual with higher fitness value will get more chance to outlet temperature. (2) Compute the logarithmic temperature
[Link] algorithm is terminated either after a certain mean difference between flue gas and [Link] data needed
number of generations or when the optimal solution has been for these computation are: temperature difference of flue gas
found. entering economizer zone and water leaving economizer(ΔE, ),
temperature difference of flue gas leaving economizer zone
IV. RESULT AND DISCUSSION and water entering economizer(ΔEF ). (3) Compute LTMD
This section will show the bond graph model of economizer using the following relationship:ΔEG = ΔE, − ΔEF /2,3 ×
and the resultingstate [Link] simulation results for step log(ΔEF /ΔEF ).(4)Compute the overall heat transfer coefficient
input changing and random disturbances are also discussed. (U) by using the following formula:N = (ℎ − ℎ ) =
?@ΔEG . This equation makes use of mass flow rate, m,
A. The governing equations enthalpy of water at the inlet and outlet of economizer (h1and
Fig.4 shows bond graph model of economizer. The C on the h2), and economizer’s surface area, A.
figure represents energy accumulator element. This type of The performance test data will be used to calculate Uand
element is viewed as enthalpy flow storage and serves to simulate the model. These data are presented inTable 1 below.
satisfy the energy conservation [Link] can be directly noticed
from Fig.4., thateffort variable, , is computed from TABLE 1
accumulation of upstream enthalpy flow of economizer plus BOILER PERFORMANCE TEST DATA
heat flow from flue gas minus downstream enthalpy flow of Performancedata I II III
economizer divided by a constant, C, (1). The constant C is a Steam load, % 110 80 50
heat capacity which equal to × × . The and are Fuel consumed, kg/hr 8400 7630 3810
specific mass of water (in. ) and specific heat of water
(in. . ) at temperature , and is economizer’s Forced draft fan out, kg/hr 129500 117600 88200
volume (in ). The ℎ is also temperature dependent, and by BFW, kg/hr 127368 115789 57895
the following relations = × = ! ", (1) can be Gas temperature, oC
rewritten as in (2). The heat transferred from flue gas to Economizer in 340 333 301
working fluid is influenced by overall heat transfer coefficient Economizer out 188 185 172
(Uin# $ ), economizer’s surface area (A in ), and Superheater in 1030 1000 830
temperature difference between flue gas and water inside Boiler bank in 820 800 640
economizer, as denoted by (3). Water/steam temperature
The flow storage element, C-field, has been identified as a Eco. in, 0C 160 160 160
differential equation where temperature output of economizer, Eco. out, 0C 208 206 201
temperature of flue gas, and boiler feed water (BFW) flow- Economizer data
rate are regarded as states and inputs, [Link] Surface, m2 547
governing equation is treated as a lumped parameter model as Volume, m3 2.99
shown below.
= & !'ℎ ( ) + +',
− 'ℎ ( )" /0
The enthalpy of water at corresponding temperature are
% - (1) computed using MATLAB function XSTEAMwhich is found
1 7'899:;5<
= 6 + ' ( − )>
in [5].
23 45 =3 (2)
The simulation that runs using calculated U from LTMD
1
method are as follows. Fig. 4-6 shows temperature responses
= 6 ?@A B − C + ' ( − )> (3)
23 45 =3 when the economizer experience step changes of flue gas
temperature.
B. Finding overall heat transfer coefficient (U)
The attempt to find overall heat transfer coefficient (U) has
been done in two ways, first using logarithmic temperature
mean difference (LTMD) and the second isby making use of
polynomial that relates boiler feed water flow-rate, fuel flow-
rate, and combustion air supply. The coefficients in the
polynomial are estimated by genetic algorithm (GA)
technique.
Finding U using LTMD is achievedthrough the following
stages: (1) Determine the enthalpy of water at the inlet and
outlet of [Link] data needed for this purpose are:
economizer’s surface area, boiler feed water flow rate, boiler
126701-5858 IJET-IJENS @ February 2012 IJENS
IJENS
International Journal of Engineering & Technology IJET
IJET-IJENS Vol: 12 No:
No 01 60
constraint function into GA, (4) running
r GA to find the
desired solution.
Here, U is estimated using a polynomial as indicated in (4)
?Y = Z × '[\] + ^ × '_`- + × ', b (4)
Then the heat which is transferred from flue gas into
working fluid is computed by (5).
NB,Y→, b = ?Y @=c A _-` − , b C (5)
The following graphs depict the reliability of the proposed
model, which employs GA for U estimation, to imitate
dynamic responses of economizer’s temperature.
Fig. 5. Step response of economizer’soutlet
’soutlet temperature at 110% of
MCC(? = 39,25 # $ , = 232,52 ° )
Fig. 8. Step response of economizer’soutlet
’soutlet temperature at 110% of MCC
(? = 20,28 # $ , = 207,79 ° )
MCC(? =
Fig. 6. Step response of economizer’soutlet temperature at 80% of MCC
36,51 # $ , = 230,77 ° )
Fig. 9. Step response of economizer’soutlet
’soutlet temperature at 80% of MCC
(? = 18,43 # $ , = 206 ° )
Fig. 7. Step response of economizer’soutlet temperature at 50% of MCC(? =
24,54 # $ , = 228,12 ° )
The simulation results show that there are discrepancies
between steady state and performance test [Link] are
probably caused by an error in economizer surface [Link]
possible sources of error are in the estimation
imation of overall heat
transfer coefficient and economizer volume.
On the other hand, estimation of U using GA has been
performed through the following phases: (1) ddetermining cost
function that will be minimized by GA, (2) ddetermining
constraint function, (3) implementing
mplementing cost function and
126701-5858 IJET-IJENS @ February 2012 IJENS
IJENS
International Journal of Engineering & Technology IJET
IJET-IJENS Vol: 12 No:
No 01 61
REFERENCES
[1] Astrom, K. J.. and Bell, R. (2000), “Drum Boiler Dynamics”,
Automatica, 36, pp 362-378.
[2] Cheres, E., (1990), “Small and Medium Size Drum Boiler Models
Suitable for Long Term Dynamic Response”, IEEE Transactions on
Energy Conversion,, Vol. 5 No. 4, pp 686-692.
686
[3] Cori, R. and d Busi, T. (1977), “Parameter Identification of a Drum Boiler
Power Plant”, Proc. 3rd Power Plant Dynamic,
Dynamic Control and Testing
Symposium,, Knoxwille, Tennessee, September 7-0, 7 1977
[4] de Mello, F. P., (1991), “Boiler Models for System Dynamic
Performance Studies”, IEEE Transactions on Power Systems,Systems Vol. 6,
No. 1pp 753-761.
[5] Holmgreen, M., 2006, XSTEAM for MATLAB, [Link], www.x
accessed at 30/10/2009.
[6] Kalogirou, S.A., 2003, Artificial intelligence for the modeling and
control of combustion processes: a review,
review Progress in Energy and
Combustion Science 29 (2003) 515–566,566, Pergamon.
Fig. 10. Step response of economizer’soutlet temperature at 50% of MCC
(? = 10,57 # $ , = 201,53 ° )
[7] McDonald, J.P. and Kwatny, H.G. (1970), “A Mathematical Model for
Reheat Boiler-Turbine-Generator
Generator Systems”, in Procs. of IEEE, PEE
Winter Power Meeting,, New York, Paper 70 CP221-PWR.
CP221
The above simulations are performed using polynomial as [8] Medjaher, K., Samantaray, A.K., Ould Bouamama, B., (2009), Bond
suggestedin (6) where the coefficients are computed using GA graph model of a vertical U-tube steam condenser coupled with a heat
exchanger, Simulation Modeling Practice and Theory,
Theory 17 (2009) 228-
Toolbox in MATLAB with parameters indicated in Table 2. 239, Elsevier.
[9] Nazaruddin, Y.Y. and Kurniawan, I., (2001), Control of Neutralization
?Y = 0,3798'[\] + 0,3748'_`- + 0.166
166', b (6) Process by an Adaptive Neuro-Fuzzy Fuzzy Controller with Genetic
Algorithm, In: Proc. Of The Third International Conference on Control
Theory and Application, p. 459-463,
463, Pretoria, South Africa
Afric
TABLE 2 [10] Nazaruddin, Y.Y. and Maulana, F., (2002), Optimizing the performance
SIMULATION PARAMETERS IN MATLAB of predictive control using genetic algorithm, In: Proc. Of The IASTED
Numerical Runge-Kutta 45 Int. Conf. Intelligent System and Control,
Control p.233-238
[11] Priandoko, B., Achmad, A.S., Nazaruddin, Y.Y., Sasongko, D., Genetic
integration method Algorithm-based
based Optimization in Nonlinear Predictive Control for
Binary Distillation Unit, Proceed. the 3rd International Conference on
Genetic Algorithm parameters Computation, Intelligence, Robotics and Autonomous System
(CIRAS’2005), Singapore, December 13-16,
13 2005.
Population 50-100 [12] Samia, L., Mabrouk, K., Mohammded, M., (2006), Modeling and
Generation 50-75 simulation of the distillation column, Asian Journal of Information
Mutation rate 0,3 Technology 5(8): 823-828, 2006.
Migration forward, fraction:: 0,2, interval: 20
Selection Uniform stochastic A.N. Aziz was born in Cilacap, Indonesia, on June 16, 1974.
1974
He is a lecturer of Physics Study Program
Progra at Jenderal
Steam Table X-Steam
Steam for MATLAB Soedirman University, Indonesia, since 1999. He has interest in
physical modeling and simulation of complex systems,
including industrial and environmental subsystems. He obtains
In equation (5), _-`B,Y is not explicitly available except his Ph.D. (Application of physical modeling and neural network technique in
boiler control systems)) in 2011 from Institut Teknologi Bandung, Indonesia.
from Boiler Performance Test Data. Despite a remarkable He is also a member of Indonesian Physical Society.
suitability of at steady state, it is not possible to employ the
proposed model since _-`B,Y is not a measured variable when
the boiler is [Link] proposed solution is roughly
estimated through its correlation with fuel flow data from
Table 1.
V. CONCLUSION
This paper has presented a methodology for improving the
performance of economizer temperature model using bond
graphh and genetic [Link] temperature behavior
is described by one order lumped parameter model plus
estimated parameters.A solutions for the equation is found in
MATLAB-Simulink
Simulink [Link] models are structured
as state space S-function-form of nonlinear class which are
solved using numerical integration sub-routine
routine provided in the
[Link] potentials of the proposed method to improve
model performance are shown from simulation result compare
to the designed data.
126701-5858 IJET-IJENS @ February 2012 IJENS
IJENS